Improving Accuracy of Air Pollution Prediction by Two Step Outlier Detection

Air pollution is one of the major problems being faced by the world today. Globally, air pollution causes more than 10 million deaths annually either directly or indirectly. The efforts to fight air pollution rely very heavily on the data that is collected from the surface sensors. Any strategy to predict or counter the effects of specific pollutants is affected by the fact that the data collected is often mixed with impurities in the form of outliers. The outliers need to be detected and removed before the data can be used for decision making. The methods used for outlier detection are limited in their accuracy and are often plagued by problems of false positives, which are readings that are falsely identified as outliers, but which in reality are novelty readings. We propose a two-step method to detect the outliers that reduces instances of false positives by differentiating true outliers from circumstantial outliers. This would help in effective prediction of air pollution or pollutant levels thereby assisting in devising effective countermeasures.